Correlation designs

advertisement
www.studyguide.pk
Correlation designs and using the Spearman’s Rank statistical test
Psychodynamic Approach
Correlation designs
As part of the methodology for the Cognitive Approach, we learned of the three experimental designs: independent
groups, repeated measures and matched pairs (see M5 Experimental Design). A further design, although not an
experimental design, is the correlation design. This design of study involves comparing two different sets of data for the
same set of participants. Each participant does two measures, and those measures are recorded and compared. This
comparison is done by testing the relationship between the two sets of results statistically.
The relationships correlation designs might identify are positive correlations (i.e. as one measure goes up, the other
measure also goes up) and negative correlations (i.e. as one measure goes up, the other measure goes down). When
there is no relationship identified, the term ‘no correlation’ is used.
The data used for a correlation design must be numerical, and both the measures must come from the same
participant. There is no independent variable and no dependent variable. There are just two variables, none of more
significance than the other. The hypothesis of such a design will not be about a ‘difference between’ but will be
hypothesising a relationship between the two measures.
Strengths of correlation designs
Weaknesses of correlation designs
good for finding relationships at the start of an
investigation; also unexpected relationships; once two
sets of data have been collected from the same
participants, a relationship can be statistically tested
the data yielded is more secure, as there are no
participant variables to affect it
the findings only show a relationship between those sets
of data, not a definite connection, it does not allow room
for the concepts of chance or another factor causing the
relationship to arise
if the data are artificial or unconnected, it is not valid
Scattergraphs
Correlation data are normally displayed graphically using a scattergraph. The scores for each measure for each
participant are used along the x-axis and the y-axis to plot a point for each participant, and then a line of best fit may be
drawn to test for a correlation.
What is referred to as the ‘eyeball test’ is used to see if there is a correlation in a scattergraph. Simply looking at the
results plotted on the chart and seeing the line of best fit should tell you if there is a correlation or not. A good measure
is to compare the number of scores on or close to the line of best fit with the number of anomalous values. The more
values there are on the scattergraph which don’t seem to fit, the less likely there is to actually be a correlation.
Spearman’s Rank correlation coefficient
The Spearman’s Rank test (Spearman’s Rank correlation coefficient) is a statistical test used to mathematically calculate
if there is a relationship between two sets of data. We will use the example of IQ and income of some participants.
Participant
IQ
Income (£)
1
2
3
4
5
6
7
8
118
103
98
124
109
115
130
110
35,000
12,000
10,000
18,000
15,000
20,000
30,000
12,000
In our study, eight participants were given an IQ test and their scores
were recorded. The participants then had their relative salaries noted.
The following instructions outline how to use the Spearman’s test.
Step 1: Rank the first variable, the lowest rank for the lowest score
In this example, this means ranking the lowest IQ with a 1 and the
highest IQ with an 8. When two participants share the same score,
take an average rank. For example, if three of them have an IQ of 100,
occupying ranks 3, 4 and 5, allocate all of them the rank 4 (average)
www.aspsychology101.wordpress.com
www.studyguide.pk
Step 2: Rank the second variable, the same way as the first
Simply rank the incomes with the lowest income receiving a 1 and the highest receiving an 8
Participant
IQ
Income (£)
1
2
3
4
5
6
7
8
118
103
98
124
109
115
130
110
35,000
12,000
10,000
18,000
15,000
20,000
30,000
12,000
IQ rank
(Step 1)
6
2
1
7
3
5
8
4
Income rank
(Step 2)
8
2.5
1
5
4
6
7
2.5
Step 3: Calculate the difference between the ranks for each participant
Take the value of Rank 2 away from the value of Rank 1 (in this example take income rank from IQ rank). This will give
the difference between the two ranks. If the resultant number is negative, don’t forget to record the minus sign
Step 4: Square the differences between the ranks
Square the value you obtain for each participant from Step 3
Step 5: Total the figures from Step 4
Find the total by adding up all the values obtained from Step 4. The Greek letter sigma, Σ, is used to show add total
Participant
IQ
Income (£)
1
2
3
4
5
6
7
8
118
103
98
124
109
115
130
110
35,000
12,000
10,000
18,000
15,000
20,000
30,000
12,000
IQ rank
(Step 1)
6
2
1
7
3
5
8
4
Income rank
(Step 2)
8
2.5
1
5
4
6
7
2.5
IQ rank –
income rank
-2
-0.5
0
2
-1
-1
1
1.5
Total (Σ) =
(IQ rank – income rank)
2
4
0.25
0
4
1
1
1
2.25
13.5
Step 6: Multiply the value of Step 5 by 6
In our example, 13.5 x 6 = 81
Step 7: Find the value of N
N is the number of pairs of observations you have, so this will simply be the number of participants, in our case 8
Step 8: Square the value of N and subtract 1 from that number
In our example, 8 x 8 = 64 and 64 – 1 = 63
Step 9: Multiply the number from Step 8 by N
In our example, 63 x 8 = 504
Step 10: Divide the value of Step 6 by the value of Step 9
In our example, 81 ÷ 504 = 0.160714285
Step 11: Calculate rho, the result of the Spearman’s test, by doing: 1 – Step 10
In our example 1 – 0.160714285 = 0.839285714
A statistical table can then be used to see if there is a correlation between the two sets of data. Statistical tables cannot
be simply generated by thinking about them, they take hundreds of mathematical studies to calculate. They are
published in masses in books so people can refer to them in order to see if there is a correlation in their data.
www.aspsychology101.wordpress.com
www.studyguide.pk
The table below is an extract from a statistical table for rho:
Level of significance for one-tailed test
N=
4
5
6
7
8
9
10
11
12
0.05
0.025
0.01
0.005
1.000
0.900
0.829
0.714
0.643
0.600
0.564
0.536
0.503
1.000
0.886
0.786
0.738
0.700
0.648
0.618
0.587
1.000
0.943
0.893
0.833
0.783
0.745
0.709
0.671
1.000
0.929
0.881
0.833
0.794
0.755
0.727
You will come across levels of
significance in more detail in M13
Inferential Statistics – MannWhitney U Test, but for now just
bear in mind that choosing a level
of significance basically means
choosing an acceptable level at
which you can reject the null
hypothesis and that the results
are due to chance, and accept the
alternative hypothesis.
Using our example study, we have a rho value of 0.893285714. We used 8 participants in the study, giving us an N value
of 8 so that row has been highlighted in the table.
The critical value shown for the appropriate level of significance in the table has to be LESS THAN rho for the hypothesis
to be proven. If the rho value is less than the critical value in the table, the null hypothesis is retained, as it cannot be
rejected. This means your rho value must be MORE THAN the value in the table (called the critical value)
So back to our example. We can definitely accept the alternative hypothesis as proven for significance of 0.05 (given
that 0.643 < 0.893). This means we can say that the results obtained showing a correlation are less than or equal to five
per cent due to chance. We can also say this for 0.025, 0.01 and 0.005 as the rho value is larger than all of the critical
values for N = 8. Because we can do so for a significance level of 0.005, we can actually say that the relationship is less
than or equal to half a per cent due to chance.
www.aspsychology101.wordpress.com
Download